Skip to content

wileland/AE

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

969 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Echo Doj0 — Train Your Emotional Intelligence

A voice-first journaling ecosystem that gamifies emotional intelligence training through receipts-grounded AI reflection.

What makes it different

Every AI insight is traceable to the user's own words. No hallucination. No projection. No "you tend to." The system reflects. It does not prescribe.

  • Provenance is Sacred — every receipt hashes to a real MemoryChunk
  • Mirror not Oracle — the system reflects your words, never diagnoses
  • Determinism Über Alles — same input, same output, always
  • 2,347 MemoryChunks — ingested from real conversation history
  • Receipts drawer — every AI response shows its source quotes

Architecture

Voice → Whisper → Entry → Chunker → Embedder → MongoDB Atlas
                                                (vector search)
                                                      ↓
Chat Message → Retrieval → Receipts (verified quotes + hashes)
                                                      ↓
                                       Companion Chat response
                                                      ↓
                                    Persona Arena (lens response)

Tech Stack

Layer Technology
Frontend React + Apollo Client + Tailwind CSS
Backend Node.js + Express + GraphQL
Database MongoDB Atlas (vector search)
Auth Supabase
AI OpenAI GPT-4o-mini + Whisper
Observability Langfuse
Deployment Render

Three Principles (enforced in code, not just philosophy)

  1. Provenance is Sacred — every insight traces to real user words via SHA-256 hash
  2. Mirror not Oracle — reflect, never predict or diagnose
  3. Determinism Über Alles — temperature: 0, stable ordering, reproducible outputs

Demo

[Live URL here]

The receipts moment:

  1. User sends a message in Companion Chat
  2. 2 receipts surface from the archive — real quotes, verified
  3. Three action chips appear: "Open Arena" / "Build Context Pack" / "Invoke a Lens"
  4. User clicks "Invoke a Lens" → Arena → selects "The Witness"
  5. Types: "I keep abandoning projects I care about."
  6. Lens response appears — grounded in the persona's defined tradition
  7. Badge: "Grounded in your archive"

Local Setup

# 1. Clone and install
git clone https://github.com/wileland/AE
pnpm install

# 2. Configure env vars
cp server/.env.example server/.env
cp client/.env.example client/.env
# Fill in MONGODB_URI, SUPABASE_JWT_SECRET, OPENAI_API_KEY, SUPABASE_URL

# 3. Seed demo personas
node server/scripts/seedPersonas.js

# 4. Start
pnpm -C server dev
pnpm -C client dev

Phases Shipped

Phase What shipped
4 Sovereign Archive — 2,347 MemoryChunks ingested
4.5 Agent Runner + inline pipeline
5 Oracle Activation — receipts-first assembly line
6 Context Pack + Companion Chat backend
6.5 Companion Chat UI — receipts drawer
7 Persona Arena — suggestedActions + invokePersona lens
8 Portfolio Prod hardening

Testing

pnpm -C server test   # 452+ backend tests
pnpm -C client test   # 76+ frontend tests
node scripts/graphql_validate_ci.mjs

About

Ashlight Echolog – MERN/Apollo journaling app

Resources

License

Contributing

Stars

Watchers

Forks

Packages

 
 
 

Contributors